My thought right now is that you had more than 100 connections to the gridftp
server:
service gsiftp
{
instances = 100
when RFT tried to form the connection for the delete. The cache of connections for delete
operations and the cache of connections for the transfer handles are different. So if there were
100 cached in transfer and RFT went to form a connection for the delete xinetd would reject it. In
this case RFT *should* retry the delete after some time, but I am not certain that it actually does.
I will dig through some code to see if my massive speculation has any merit.
